Sendmail Delay?
 

Hi Folks.

I have an e-mail distribution list that is stored in a database table, and users can "sign up" and have themselves added to the list.

Every week, before sending out the notification I have to send out, I run a little perl script to extract the list from the database, and output it to a text file that is listed in /etc/aliases.

Then I simply send the message to the alias and off it goes.

Recently I moved this web site (etc.) to a new server. It had been on a machine running RHEL4. The new machine runs RHEL5. I didn't make any configuration changes to any of the sendmail files, and as far as I can remember, I haven't fiddled with anything that should have messed with the way mail is handled.

However... On the old machine, it used to take sendmail about an hour or so to send out all the messages (I think it first takes the alias list and then re-orders it by domain names, probably so that it can send all the foo.com addresses at once?). There are probably 3,000 addresses.

On the new machine, the first time I sent a message to the alias, it took over 24 hours for it to send all the messages out. I can live with an hour--24 hours seems too long.

I know this is a TOTAL stretch, but does anyone have the slightest idea what might have changed from RHEL4 to RHEL5 that would cause this slowdown? I looked through configuration files for sendmail on both machines and they look identical.

I looked for a "delay between messages" setting or some such and was unable to find anything similar.

Again, I know that this is a stretch, but figured it couldn't hurt to ask.

Thanks in advance for any replies. :)

G.--

Any further clues in /var/log/maillog?

Well I just sent the notification five or six minutes ago, so the log just has the initial verification that the message was sent to the list alias:

Nov 8 12:36:06 myserver sendmail[10819]: lA8Ia6ie010819: to=myemail-list , ctladdr=webmaster@xxxxx.edu (48/48), delay=00:00:00, xdelay=00:00:00, mailer=relay, pri=32400, relay=[127.0.0.1] [127.0.0.1], dsn=2.0.0, stat=Sent (lA8Ia6a3010820 Message accepted for delivery)

(I changed the name of "myserver" and "myemail-list" and "webmaster@"...)

The box is pretty quiet at this point. I have one CPU at ~25% or so, and the other 7 are all at 0.0% or are near that (dual Xeon processors).

Unfortunately, until I get my e-mail message from the system, I won't be able to tell you how long this took (and the domain I'm in is close to the end of the alphabet, so I usually get mine close to the end of the process....)

G.--

Well it's been a couple of hours, and we're up to the "Ks."

Nov 8 14:52:16 foo sendmail[10821]: STARTTLS=client, relay=koxxxx.com.inbound10.xxx.net., version=TLSv1/SSLv3, verify=FAIL, cipher=AES256-SHA, bits=256/256
Nov 8 14:52:18 foo sendmail[10821]: lA8Ia6a3010820: to=mxxxxx@koxxxxx.com, delay=02:16:12, xdelay=00:00:14, mailer=esmtp, pri=87632771, relay=koxxxxx.com.inbound10.xxxxx.net. [208.65.144.2], dsn=2.0.0, stat=Sent (Backend Replied [10773374.2493070256.005.00-005.p01c11m015.mxlogic.net]: .0.0 lA8KqHqb009864 Messag (Mode: normal))
Nov 8 14:52:18 foo sendmail[10821]: lA8Ia6a3010820: to=kxxx@krxxxx.com, delay=02:16:12, xdelay=00:00:00, mailer=esmtp, pri=87632771, relay=mx1.xxxxx.com. [66.216.121.101], dsn=2.0.0, stat=Sent (Ok: queued as ABD451B615F)

It almost looks like when it get stuck on one it sits and doesn't do anything else--is this correct?

Nov 8 15:00:03 foo sendmail[10821]: lA8Ia6a3010820: timeout waiting for input from xxx.bar.com. during client greeting
Nov 8 15:05:06 foo sendmail[10821]: lA8Ia6a3010820: timeout waiting for input from xxx.bar1.com. during client greeting

I guess I was under the impression that this thing would have multiple threads sending out multiple messages at a time and not traveling along in just a linear trek through the list of addresses? Does that make sense?

G.--

By default sendmail sends out a queue runner every 5 minutes. The default Timeout.hoststatus is 30 minutes, which means if a timeout occurs attempting a delivery (a temporary failure) that host will be ignored for 30 minutes, even though queue runners are being sent every 5 minutes. You might try playing with the Timeout.hoststatus settings - confTO_HOSTSTATUS in sendmail.mc and send out a couple of queue runners to see if you can clear the queue more quickly.

Do you know what that TLS verify=fail is all about?

I don't know what the TLS verify=fail thing is. Is a receiving e-mail server requiring that the message be encrypted or some such before delivery?

It looks like it finished up sending messages to legitimate addresses around 4:00 yesterday afternoon. I'll take three and a half hours--that's sure better than 23 or 24 hours like it was a month ago.

When looking through the logs, it appears that it spends a lot of time re-sending to possibly bad addresses. I'll do some clean-up next week and get rid of addresses that I think are no longer active, and see if that helps speed it up any further.

Thanks for the help!

G.--